Research On Image Classification Algorithm Based On Conceptual Cognition

Concept cognitive learning is a hot research field of cognitive science.Recently,application research based on concept cognitive learning has become a significant topic.However,there still exist many problems in the process of applying the concept cognitive learning to image classification tasks,mainly including 1)How to transform the image data into the formal context;2)How to apply the theory of the concept cognitive to the specific image classification tasks.In response to the above problems,the thesis combines concept cognition and decision tree ideas,proposes a image classification method based on concept cognitive,named as CCIC(Concept Cognitive Image Classification).The proposed method is composed of the global image formal context construction algorithm(IFCC)and the progressive concept cognitive image classification algorithm based on object-oriented concepts(POCCIC).The main research production of the thesis is as follows:1)A global image formal context construction algorithm is proposed.Firstly,SIFT technology is used to extract the image features,and PCA is used to reduce the dimensionality of features to generate an image feature matrix,and then,bag of visual word model is adopted to convert the image feature matrix into an image feature distribution matrix.Finally,the feature distribution matrix is transformed into a binary image formal context according to the global feature distribution threshold,which can achieve the consistent conversion of the image formal context.2)A progressive concept cognitive image classification algorithm based on object-oriented concept is proposed.Firstly,inspired by the progressive concept cognitive model,a theoretical model of progressive concept cognitive based on object-oriented concepts is given,which gradually learns the approximate object-oriented concept of images from the basic attributes of the object;then,draws on the idea of decision trees and uses The tree structure organizes the cognitive results and gives the update conditions;finally,it is applied to the classification task.3)Experimental verification analysis.Experiment with the method in the public data set,and get the optimal external conditions of the method with the aid of ablation experiments.Comparative experiments show that when the upper limit of concept connotation is 16 and the upper limit of visual word capacity is 1000,compared with the decision tree model,the optimal classification accuracy of this method is increased by 5 percentage points;at the same time,in the comparison of TOP 5 classification accuracy,the classification effect is excellent In the image classification method based on multi-level concept lattice,the maximum increase is about 10%.This paper proposes a image classification method.The experimental results show the effectiveness of the method and can be applied to image classification tasks.The work of this paper further broadens the concept of cognitive theory and its application,which has certain academic significance and application value.